The RSE-BERT-large-Transfer is trained with 2 relations including:

1. entailment
2. paraphrase

The BERT-large-uncased model is used as initialization.

It can be used ideally for Transfer datasets - (Downstream Tasks).